TDA2030 DC at output

[MrStab]

hi guys,

i decided i'd been away from the forum too long last night, and good thing i did, because i'm a bit stumped!

any idea why the single-supply circuit in the TDA2030 datasheet would be showing DC at the output (after the 2,200uF cap)?



i can't find any shorts or solder bridges, i've tried replacing the cap and the circuit itself seems fine. there's a chance i've fried the chip from accidentally soldering the input cap to pin 2 instead of pin 1, but i find it difficult to believe that would allow any DC past the output cap. or would it? the chip gets really hot when a load is applied and it's prone to oscillate, so maybe it is dead, but i'd prefer to solve this problem before using the replacement. the TDA2030 is heatsinked with thermal paste etc. (though it could be better), and there's no enclosure to speak of yet (so no V-/tab problems)

i'm using a 19V supply for now. i measure about 6V on the output when no load is connected, which i think goes up to 9V when a speaker is connected (though that increase may be me mixing up tests in my head, it's too late at night to double-check). either way it's at 6V idle.

FWIW, the input is fed from a basic non-inverting opamp stage with some gain and an attenuator on the output, then coupled into the power amp stage. all that works fine. the negative lead from the speaker cable follows its own path to star ground, everything else grounded shares the same line as the TDA2030.

cheers for any thoughts! it's not a critical project, but it would be a personal victory to finally get this kind of chip-amp working after i gave up on one a few years ago.

[PRR]

> about 6V on the output when no load is connected

That is correct. How can a dead-end cap know you want it to go to zero? You need to tell it by tying something to zero. If afraid to risk your speaker, try 220 Ohms to ground. (Very high R values won't pull-down well because the huge e-cap leaks a bit.)

> which i think goes up to 9V when a speaker is connected

Well, that is just wrong.

Get some sleep.

Is the output cap polarity correct? If backward, it will conduct like crazy, heat the chip and the speaker.

[MrStab]

it's late for society, not for my chronotype!

i'm certain that there's DC ranging from 6-9V with the speaker connected, just wasn't sure about the increase. fair point about the unloaded voltage, seems a bit silly in hindsight - it's floating!

200R measures 0V across. weird. i'll need to have a look at the housing (a Heil Talk Box) and the surroundings, maybe there's something offboard i'm overlooking. the cathode is connected to the speaker.

thanks for helping me narrow it down, i'll report back

[anotherjim]

Could be the chip amp oscillating at supersonic. Those caps on +Vs are meant to be there and close to the chips power pins. 100uF is the small "local" cap - it expects there is a decent sized one on the PSU output, 1000uF or bigger.

But as Paul said, the usual is a leaky output cap - are you measuring that DC on the speaker? Is your speaker voice coil still intact?

[MrStab]

yknow, i thought that seemed a bit low. i have 220uF across the power supply rails atm, an additional 100uF/100n was close to the chip but has since been rerouted so maybe it needs to be closer again. could lacking that reservoir maybe be the cause of this problem then?

i'm getting into standard chip-amp troubleshooting which i should just UTFSE for, admittedly. at first i just thought i'd come across some obscure issue.

i've tried two brand new 2,200uF caps. the first one i removed measures 0.9% Vloss and otherwise tests fine. FWIW, i made one parts substitution: as per the GeneralGuitarGadgets amp, the zobel network is 4.7R/100n as opposed to 1R/220n. this was because i had no 1R resistors of a decent rating. i assumed this wouldn't be a problem, but maybe it is?

cheers Jim! i'll poke around at it once coffee #2 has kicked in. i'll check the speaker enclosure for something i'm overlooking and try sorting out the caps.

[anotherjim]

My experience is the Zobel isn't that fussy. At least I've gotten away with similar substitution trick so far. If I don't have a value, I find another one to copy that I can make.

Single supply, you don't have to worry about tab insulation -  it is connected to pin3 the -Vs supply I think. Unless for some reason you want the -v supply floating apart from ground.

It's a big daddy op-amp really, so if you have the same test situation -  +V/2 off the 100k dividers should be found on the output pin and those electro caps on the inputs must also NOT be leaking.

[MrStab]

vast improvement, thanks! i used the original 2,200uF cap across the PS rails, put 100uF right across the chip and it's mostly working as expected. still some distortion and trebbliness, but i'm sure there are a million posts out there with solutions. heatsink is getting a bit warm, it's not the biggest to begin with but it's definitely not boiling anymore. could this slightly-sub-par performance be due to IC damage? it's still the absolute last thing i wanna swap out.

i've just reconfirmed that the speaker itself is fine. it doesn't distort and the output covers more of the spectrum. the PSU i'm using is slightly noisy, but i'm accounting for that - i want to know this is viable before i buy a decent adapter. i've only built one proper PSU with transformers etc. for personal use, but this is for a friend and he's not a politician so i don't want to put him at risk.

Single supply, you don't have to worry about tab insulation

i had a feeling and was going to ask! thanks for clearing that up. that drastically increases the size of my heatsink.

edit: i realise i haven't measured sag or anything, i'll get on that.

[anotherjim]

Cool 
I think you can trust the 2030's built in protection + the overshoot diodes on the output.  That chips been around a long time, outliving many rivals.

I built one myself not long ago, but as I had a 9-0-9 traffy to hand I went for the dual supply version - saved the big output cap but had to insulate the heatsink. I don't work it too hard with an 8ohm speaker. It do get warm, but nothing serious.

One thing I did notice, was the data sheet gain setting is generous - more than enough for any line level input or an amp sim pedal, so I cut the gain by reducing the feedback R. You have x30, but you can probably reduce to x15. Less gain should improve stability a bit too.

[MrStab]

reassuring to know it's a sturdy chip. it's for talkbox use, but i'd like to have the option of an external output. halving the input (as opposed to the TDA's gain) suggests a reduction to 15 would be fine for this purpose. could the distortion be caused by the quality reduction i've read about by using the TDA20** in single-supply mode, with the big output cap? maybe a reduction in gain could mitigate this effect?

no sag at all on either the input opamp or the power amp stage. i changed the opamp from a (definitely-counterfeit!) 4558 to an OPA2107, but the problem's still there. my pickups have quite a high output, but the voltage range should be able to handle that. speaker is 8 ohms, and i think the "Do not exceed 100W peak" sticker on the enclosure suggests i'm safe with that. lol
seems like a bit of a nitpicky minor detail i need to fix, as the circuit works overall.

(i have to get rid of these random opamps somehow!)

[MrStab]

making the opamp input stage unity-gain (and below) and lowering the gain by varying degrees on the TDA2030 can reduce the distortion at disappointing levels. that (and maybe the apparent low-end rolloff) make me think something's just not functioning right.

it's only upon hard strums that the distortion is a problem, and it's not necessarily an ugly distortion, just unwanted. i can't really find any wrong component values in the zobel network, power amp feedback loop LPF or input caps+resistors that would be causing the trebbly sound. which isn't obnoxiously-bad either, but it's still not ideal. it is a SMPS for now, but as i say, no sag that i can detect. the volume trimpot between the opamp input stage and power amp does get noisy from PSU whine as it shifts towards Vref, though.

maybe a limitation of the IC at 18V? i can't spot anything on the datasheet - no slopes on the Distortion vs. Output Power charts for ~3W, and no way is the signal exceeding input voltage. i put a 100nF ceramic across the 100uF on the power amp for good measure while i was in there.

i could draw up a schem but it's just a really basic opamp gain stage, volume pot to Vref straight out of the output then 1uF cap into the TDA2030 datasheet diagram.

[anotherjim]

Re volume control - and "shifts toward Vref". Should there be any Vref there? Have you got a cap between op-amp and volume pot? If the pot is 22k, the cap from op-amp probably needs to be at least 100nF.
Try giving the + supply to the input op-amp its own RC filter to isolate from amp/psu ripple. There must be some reaction there when the amp is working hard, although I know you say you can't see any.
Probably you can use the 2030 Vref for the op-amp, but it's own might be safer.

[MrStab]

ok, i'll try decoupling the input opamp with 47R/100uF or something. FWIW, i've seen measurable sag via. this DMM on other pedals which were inaudible, but it's not exactly lightning-fast and i'm only getting 1 decimal place, so it's not infallible.

here's what i have, signal-wise:



by my vague and useless language like "shifts toward Vref", i meant as the wiper moves towards Vref. lol
i usually prefer to couple volume pots with a 10uF & 1uF cap when they're gonna be on the panel, but i figured i could save some space by omitting one if both ends are at roughly 1/2 supply, even if there's a bit of an offset.

there's a separate Vref for the signal: currently, it's just a 10k resistor divider with a 47uF cap, but there's actually a wider circuit (for mic return path) i plan to work on later, focused around a currently-empty quad opamp socket which happens to include a buffered bias. so i could activate that to see if it helps. which just made me realise something: AFAIK, when using an unbuffered voltage divider bias, it's preferable to use a cap to ground from R6 instead of just a straight connection to Vref. i wired it straight to Vref because the final circuit will expect the buffered bias. could this be contributing? i don't know too much about that nuance. there are only maybe like 4 other bias points tops, which are all currently unloaded by the absence of the quad opamp.

there's a wee RF filter on the input, force of habit. the two HPFs throughout the circuit (C1/R4 and R5/100k pwr amp bias resistor) shouldn't account for the weak lows, and the values are correct. R3 is currently shorted. the trimpot is 100k. the 1uF into the TDA2030 is a non-polar film cap.

unfortunately i probably won't be able to work on it until monday, but i might be able to get half an hour in tomorrow afternoon.

cheers, Jim!

[Rob Strand]

200R measures 0V across. weird. i'll need to have a look at the housing

PRR's advice is right on track.  Given you get 0V there's probably no fault.

Your multimeter has a 10MEG ohm impedance, the RC time constant with a 2000uF cap is 20000 seconds or 5.6hrs.  So you have to wait a long time to see the DC voltage drop with a multimeter connected

When use the 200 ohm resistor the time constant reduces to 0.4seconds so the DC voltage stabilizes in a much shorter time (something more observable by humans!).

[anotherjim]

The op-amp + in resistor is the only one that can go straight on a voltage divider ref without a bypass cap. Here, Vref is just a DC reference level, there is no signal. The - input however, does carry signal so you should have a bypass cap on Vref OR a buffered reference. I personally don't like returning the feedback network to Vref and would fit a cap and return to ground instead. Besides, a cap there is an opportunity to limit the bass response - do you need a lot of bass coming up the mouth tube?

Also, I wouldn't return a pot to Vref without a bypass or buffered ref.
If you do return the pot to Vref, you don't need a blocking cap on the pre-amp output.

Thing is, if signal passes thru the Vref resistors, they modulate the reference voltage. That in turn feeds some signal on the op-amp + input. You don't want that; it's meant to be pure DC reference level. A large enough bypass cap on the Vref should prevent it from getting modulated by signal as it passes thru the cap (line of least resistance) rather than the resistors.

[MrStab]

thanks for the reply, Rob. problem now is just minor distortion, everything else is working as-is. in retrospect i was being kind of an idiot with the original problem. lol

i'll get the buffered Vref online then, Jim, you make a good case for it. just realised i only have R2R quad opamps, which will die above 16V, so i'll need to score some TL074s on Monday. i'll let you know how it goes.

cheers, guys!

[MrStab]

I finally got some TL074's and while the distortion was reduced somewhat by buffering the Vref, it's still a problem.

I've tried removing the volume pot and adding 100uF & 100nF caps across the opamp supply rails. R3 is still bypassed for unity gain, and the problems still worsen when it isn't (and gain goes up), regardless of the changes.

FWIW i've built and used this kind of input stage plenty, with buffered Vref from the start, and only ever ran into this kind of distortion when power supply or biasing was an issue. or if the next stage can't handle it for whatever reason, which is all i can think of atm.

the whole time i've been bypassing the 15uF cap in the Heil Talkbox, so there's just the 2,200uF cap between the output and the speaker, but i tried it with the 15uF in series. naturally a lot of lows were cut, but the distortion threshold also lowered. dunno if this is relative to the problem. should the input cap to the power amp stage be as high as 1uF?

[MrStab]

after the above changes, reducing the gain on the TDA2030 seems to help more than it did before. working on finding a good balance right now.

[anotherjim]

If you've got the Vref sorted out & it's as per the diagram above -  it should be ok - in itself.
In my experience, if larger supply caps don't fix it (and there is a limit to how big those caps can be since the PSU has to be able to feed them), it's often the power supply itself not liking the peak currents when the amp is set louder.
You are using an SMPS supply? Do you have a plain old transformer type you can try?

That 1uF input cap should be fine. However, if you get scratchy noise from turning the level pot, there could well be something wrong with it. It could just as well be 100nF film if you want to avoid polarized.

There's one reason I don't like pots on Vref -  a polarized coupling cap isn't polarized - both ends of the cap are Vref and you can't say which way around it should be.

[MrStab]

a scenario is forming in my head:

TDA2030 datasheet's stock gain is about 33, right (i've been changing it back & forth)? my slow DMM picks up something like 400mVrms out of my guitar, and this slightly-unscientific list suggests 833mV from the Sh8B pickup: http://www.seymourduncan.com/forum/showthread.php?278978-Seymour-Duncan-mV-Data (which is more or less what the friend it's for will be using).

i can't tell if they're using RMS, peak or Vpp, but in any scenario with these numbers, it now makes total sense for it to clip, even with no prior gain. so i may need to consider another IC or PSU for a little bit more headroom.

i am using an SMPS at the moment, no proper 18V linear supply unfortunately. 19V/3.4A available. the 1uF to the gain stage is non-polarised.

[MrStab]

thanks a lot for your help, Jim - you pointed out the overall problem earlier, but it was masked by another problem (which you also pointed out!).

the chipamps don't really get competitive against each other at 18V. I was hoping to be able to use 18V 1A pedal adapters for easier replacement, and avoid sticking a transformer in there with my limited skills (plus it'd probably radiate into the speaker). i'll look around for regulated 24V adapters outside the purpose of pedals, as the best i can find there is an EHX adapter with only 100mA available.